Overview

From scratch implementation of ViT

Key Features

• Transformer Architecture

Technical Details

• Framework: PyTorch
• Dataset: Custom
• Category: Computer Vision

Implementation Details

Implmented a ViT Architecture from Scratch using Pytorch on a subset of Food-101 dataset.

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

Dataset Information

Dataset (Train): Subset of Food101 (3 classes-255 images total) Dataset (Test): Subset of Food101 (3 classes-75 images total)

Frameworks

Pytorch

Results

Training loss: 1.20
Test loss: 1.52

Authors

• @YuvrajSingh

Source Code

📁 GitHub Repository: ViT

View the complete implementation, training scripts, and documentation on GitHub.